Coverage Report

Created: 2025-09-29 18:47

next uncovered line (L), next uncovered region (R), next uncovered branch (B)
/root/doris/be/src/runtime/fragment_mgr.h
Line
Count
Source
1
// Licensed to the Apache Software Foundation (ASF) under one
2
// or more contributor license agreements.  See the NOTICE file
3
// distributed with this work for additional information
4
// regarding copyright ownership.  The ASF licenses this file
5
// to you under the Apache License, Version 2.0 (the
6
// "License"); you may not use this file except in compliance
7
// with the License.  You may obtain a copy of the License at
8
//
9
//   http://www.apache.org/licenses/LICENSE-2.0
10
//
11
// Unless required by applicable law or agreed to in writing,
12
// software distributed under the License is distributed on an
13
// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
14
// KIND, either express or implied.  See the License for the
15
// specific language governing permissions and limitations
16
// under the License.
17
18
#pragma once
19
20
#include <gen_cpp/FrontendService_types.h>
21
#include <gen_cpp/QueryPlanExtra_types.h>
22
#include <gen_cpp/Types_types.h>
23
#include <gen_cpp/types.pb.h>
24
25
#include <cstdint>
26
#include <functional>
27
#include <iosfwd>
28
#include <memory>
29
#include <mutex>
30
#include <string>
31
#include <unordered_map>
32
#include <vector>
33
34
#include "common/be_mock_util.h"
35
#include "common/status.h"
36
#include "http/rest_monitor_iface.h"
37
#include "runtime/query_context.h"
38
#include "runtime_filter/runtime_filter_mgr.h"
39
#include "util/countdown_latch.h"
40
#include "util/hash_util.hpp" // IWYU pragma: keep
41
#include "util/metrics.h"
42
43
namespace butil {
44
class IOBufAsZeroCopyInputStream;
45
}
46
47
namespace doris {
48
#include "common/compile_check_begin.h"
49
extern bvar::Adder<uint64_t> g_fragment_executing_count;
50
extern bvar::Status<uint64_t> g_fragment_last_active_time;
51
52
namespace pipeline {
53
class PipelineFragmentContext;
54
} // namespace pipeline
55
class QueryContext;
56
class ExecEnv;
57
class ThreadPool;
58
class PExecPlanFragmentStartRequest;
59
class PMergeFilterRequest;
60
class RuntimeProfile;
61
class RuntimeState;
62
class TPipelineFragmentParams;
63
class TPipelineInstanceParams;
64
class TScanColumnDesc;
65
class TScanOpenParams;
66
class Thread;
67
class WorkloadQueryInfo;
68
69
std::string to_load_error_http_path(const std::string& file_name);
70
71
template <typename Key, typename Value, typename ValueType>
72
class ConcurrentContextMap {
73
public:
74
    using ApplyFunction = std::function<Status(phmap::flat_hash_map<Key, Value>&)>;
75
    ConcurrentContextMap();
76
    Value find(const Key& query_id);
77
    void insert(const Key& query_id, std::shared_ptr<ValueType>);
78
    void clear();
79
    void erase(const Key& query_id);
80
12
    size_t num_items() const {
81
12
        size_t n = 0;
82
1.53k
        for (auto& pair : _internal_map) {
83
1.53k
            std::shared_lock lock(*pair.first);
84
1.53k
            auto& map = pair.second;
85
1.53k
            n += map.size();
86
1.53k
        }
87
12
        return n;
88
12
    }
_ZNK5doris20ConcurrentContextMapINS_9TUniqueIdESt8weak_ptrINS_12QueryContextEES3_E9num_itemsEv
Line
Count
Source
80
12
    size_t num_items() const {
81
12
        size_t n = 0;
82
1.53k
        for (auto& pair : _internal_map) {
83
1.53k
            std::shared_lock lock(*pair.first);
84
1.53k
            auto& map = pair.second;
85
1.53k
            n += map.size();
86
1.53k
        }
87
12
        return n;
88
12
    }
Unexecuted instantiation: _ZNK5doris20ConcurrentContextMapISt4pairINS_9TUniqueIdEiESt10shared_ptrINS_8pipeline23PipelineFragmentContextEES6_E9num_itemsEv
89
36
    void apply(ApplyFunction&& function) {
90
4.60k
        for (auto& pair : _internal_map) {
91
            // TODO: Now only the cancel worker do the GC the _query_ctx_map. each query must
92
            // do erase the finish query unless in _query_ctx_map. Rethink the logic is ok
93
4.60k
            std::unique_lock lock(*pair.first);
94
4.60k
            static_cast<void>(function(pair.second));
95
4.60k
        }
96
36
    }
_ZN5doris20ConcurrentContextMapISt4pairINS_9TUniqueIdEiESt10shared_ptrINS_8pipeline23PipelineFragmentContextEES6_E5applyEOSt8functionIFNS_6StatusERN5phmap13flat_hash_mapIS3_S7_NSB_4HashIS3_EENSB_7EqualToIS3_EESaIS1_IKS3_S7_EEEEEE
Line
Count
Source
89
12
    void apply(ApplyFunction&& function) {
90
1.53k
        for (auto& pair : _internal_map) {
91
            // TODO: Now only the cancel worker do the GC the _query_ctx_map. each query must
92
            // do erase the finish query unless in _query_ctx_map. Rethink the logic is ok
93
1.53k
            std::unique_lock lock(*pair.first);
94
1.53k
            static_cast<void>(function(pair.second));
95
1.53k
        }
96
12
    }
_ZN5doris20ConcurrentContextMapINS_9TUniqueIdESt8weak_ptrINS_12QueryContextEES3_E5applyEOSt8functionIFNS_6StatusERN5phmap13flat_hash_mapIS1_S4_NS8_4HashIS1_EENS8_7EqualToIS1_EESaISt4pairIKS1_S4_EEEEEE
Line
Count
Source
89
24
    void apply(ApplyFunction&& function) {
90
3.07k
        for (auto& pair : _internal_map) {
91
            // TODO: Now only the cancel worker do the GC the _query_ctx_map. each query must
92
            // do erase the finish query unless in _query_ctx_map. Rethink the logic is ok
93
3.07k
            std::unique_lock lock(*pair.first);
94
3.07k
            static_cast<void>(function(pair.second));
95
3.07k
        }
96
24
    }
97
98
    Status apply_if_not_exists(const Key& query_id, std::shared_ptr<ValueType>& query_ctx,
99
                               ApplyFunction&& function);
100
101
private:
102
    // The lock should only be used to protect the structures in fragment manager. Has to be
103
    // used in a very small scope because it may dead lock. For example, if the _lock is used
104
    // in prepare stage, the call path is  prepare --> expr prepare --> may call allocator
105
    // when allocate failed, allocator may call query_is_cancelled, query is callced will also
106
    // call _lock, so that there is dead lock.
107
    std::vector<std::pair<std::unique_ptr<std::shared_mutex>, phmap::flat_hash_map<Key, Value>>>
108
            _internal_map;
109
};
110
111
// This class used to manage all the fragment execute in this instance
112
class FragmentMgr : public RestMonitorIface {
113
public:
114
    using FinishCallback = std::function<void(RuntimeState*, Status*)>;
115
116
    FragmentMgr(ExecEnv* exec_env);
117
    ~FragmentMgr() override;
118
119
    void stop();
120
121
    // execute one plan fragment
122
123
    Status exec_plan_fragment(const TPipelineFragmentParams& params, const QuerySource query_type,
124
                              const TPipelineFragmentParamsList& parent);
125
126
    void remove_pipeline_context(std::pair<TUniqueId, int> key);
127
    void remove_query_context(const TUniqueId& key);
128
129
    Status exec_plan_fragment(const TPipelineFragmentParams& params, const QuerySource query_type,
130
                              const FinishCallback& cb, const TPipelineFragmentParamsList& parent);
131
132
    Status start_query_execution(const PExecPlanFragmentStartRequest* request);
133
134
    Status trigger_pipeline_context_report(const ReportStatusRequest,
135
                                           std::shared_ptr<pipeline::PipelineFragmentContext>&&);
136
137
    // Can be used in both version.
138
    MOCK_FUNCTION void cancel_query(const TUniqueId query_id, const Status reason);
139
140
    void cancel_worker();
141
142
    void debug(std::stringstream& ss) override;
143
144
    // input: TQueryPlanInfo fragment_instance_id
145
    // output: selected_columns
146
    // execute external query, all query info are packed in TScanOpenParams
147
    Status exec_external_plan_fragment(const TScanOpenParams& params,
148
                                       const TQueryPlanInfo& t_query_plan_info,
149
                                       const TUniqueId& query_id,
150
                                       const TUniqueId& fragment_instance_id,
151
                                       std::vector<TScanColumnDesc>* selected_columns);
152
153
    Status apply_filterv2(const PPublishFilterRequestV2* request,
154
                          butil::IOBufAsZeroCopyInputStream* attach_data);
155
156
    Status merge_filter(const PMergeFilterRequest* request,
157
                        butil::IOBufAsZeroCopyInputStream* attach_data);
158
159
    Status send_filter_size(const PSendFilterSizeRequest* request);
160
161
    Status sync_filter_size(const PSyncFilterSizeRequest* request);
162
163
    std::string to_http_path(const std::string& file_name);
164
165
    void coordinator_callback(const ReportStatusRequest& req);
166
167
0
    ThreadPool* get_thread_pool() { return _thread_pool.get(); }
168
169
    // When fragment mgr is going to stop, the _stop_background_threads_latch is set to 0
170
    // and other module that use fragment mgr's thread pool should get this signal and exit.
171
0
    bool shutting_down() { return _stop_background_threads_latch.count() == 0; }
172
173
0
    int32_t running_query_num() { return cast_set<int32_t>(_query_ctx_map.num_items()); }
174
175
    std::string dump_pipeline_tasks(int64_t duration = 0);
176
    std::string dump_pipeline_tasks(TUniqueId& query_id);
177
178
    void get_runtime_query_info(std::vector<std::weak_ptr<ResourceContext>>* _resource_ctx_list);
179
180
    Status get_realtime_exec_status(const TUniqueId& query_id,
181
                                    TReportExecStatusParams* exec_status);
182
    // get the query statistics of with a given query id
183
    Status get_query_statistics(const TUniqueId& query_id, TQueryStatistics* query_stats);
184
185
    std::shared_ptr<QueryContext> get_query_ctx(const TUniqueId& query_id);
186
187
private:
188
    struct BrpcItem {
189
        TNetworkAddress network_address;
190
        std::vector<std::weak_ptr<QueryContext>> queries;
191
    };
192
193
    Status _get_or_create_query_ctx(const TPipelineFragmentParams& params,
194
                                    const TPipelineFragmentParamsList& parent,
195
                                    QuerySource query_type,
196
                                    std::shared_ptr<QueryContext>& query_ctx);
197
198
    void _check_brpc_available(const std::shared_ptr<PBackendService_Stub>& brpc_stub,
199
                               const BrpcItem& brpc_item);
200
201
    // This is input params
202
    ExecEnv* _exec_env = nullptr;
203
204
    // (QueryID, FragmentID) -> PipelineFragmentContext
205
    ConcurrentContextMap<std::pair<TUniqueId, int>,
206
                         std::shared_ptr<pipeline::PipelineFragmentContext>,
207
                         pipeline::PipelineFragmentContext>
208
            _pipeline_map;
209
210
    // query id -> QueryContext
211
    ConcurrentContextMap<TUniqueId, std::weak_ptr<QueryContext>, QueryContext> _query_ctx_map;
212
    std::unordered_map<TUniqueId, std::unordered_map<int, int64_t>> _bf_size_map;
213
214
    CountDownLatch _stop_background_threads_latch;
215
    std::shared_ptr<Thread> _cancel_thread;
216
    // This pool is used as global async task pool
217
    std::unique_ptr<ThreadPool> _thread_pool;
218
219
    std::shared_ptr<MetricEntity> _entity;
220
    UIntGauge* timeout_canceled_fragment_count = nullptr;
221
};
222
223
uint64_t get_fragment_executing_count();
224
uint64_t get_fragment_last_active_time();
225
#include "common/compile_check_end.h"
226
} // namespace doris